TerminatorNet: comprehensive identification of intrinsic transcription terminators in bacteria

<h4>Motivation</h4>The primary mechanism for transcription termination in bacteria is intrinsic terminators. These terminators influence transcript stability and play key roles in gene regulation. Existing computational methods for genome-wide terminator identification have been designed and evaluated based on a small number of experimentally evinced terminators often from only one or two organisms.<h4>Results</h4>We present TerminatorNet, a system for identifying intrinsic transcription terminators throughout bacteria. TerminatorNet uses a neural network model trained on a large set of experimentally characterized transcription terminators from a variety of bacterial genomes. TerminatorNet identifies 98% of terminators and has a false positive rate of 3%, substantially better than existing approaches. TerminatorNet commonly identifies terminators at the ends of operons. We applied TerminatorNet to thousands of genomes across the taxonomic spectrum of prokaryotes, creating a repository of tens of millions of terminators. We observe heavy use of intrinsic termination in some groups, such as Bacillota, and rare use in other groups such as archaea. We also observe a wealth of instances of DNA uptake signal sequences, important components of transformation specificity for some competent bacteria, in terminators identified in Neisseriaceae and Pasteurellaceae.<h4>Availability and implementation</h4>TerminatorNet and its repository of identifications are available for use via a webserver: https://cs.wellesley.edu/∼btjaden/TermNet. The source code is available at GitHub https://github.com/btjaden/TerminatorNet and Zenodo https://doi.org/10.5281/zenodo.18406126.